PAX2 has been cited as a technically robust biomarker which nicely delineates precancerous lesions of the endometrium when the endometrial intraepithelial neoplasia (EIN) classification scheme is used. Its utility in distinguishing between atypical and nonatypical hyperplasia when applied within the 1994 World Health Organization classification system is questionable. The purpose of this study was to evaluate PAX2 in a side by side comparison of its staining patterns in a series of endometrial samples that were classified using both systems. A total of 108 precancerous endometrial cases were identified, of which 30 cases were deemed nonhyperplastic by consensus agreement and 11 cases lost the tissue of interest on deeper sections. The remaining 67 cases were categorized according to the 1994 World Health Organization criteria and EIN scheme by 2 gynecologic pathologists. PAX2 staining was scored in lesional tissue as normal or altered (lost, increased, or decreased) compared with nonlesional background. The most common pattern of alteration was complete loss of nuclear PAX2 staining (86.3%) followed by decreased staining (11.3%) and markedly increased staining (2.3%). PAX2 alterations correlated well with EIN diagnoses (33/36, 92%) compared with benign hyperplasia (2/13, 15%) but were less useful when the 1994 World Health Organization classification system was applied (PAX2 alteration in 22/25 (88%) of atypical hyperplasia cases versus 16/25 (64%) of nonatypical hyperplasia cases). Forty-five percent of follow-up hysterectomies with a previous PAX2-altered biopsy case harbored adenocarcinoma. In conclusion, PAX2 may be a helpful adjunct stain and training tool when the features of atypical hyperplasia/EIN are in question.